Spectral Diagnostics of Emergence: A Random Matrix Baseline and a Mesoscopic Regime From Competing Scales
IEEE Access · 2026 · 14 · 83457-83470
DOI: 10.1109/access.2026.3696391 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Alain Tcheukam Siwe. Die Arbeit nennt 72 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Higher Colleges of Technology, United Arab Emirates.
Abstract (Originalsprache)
Understanding how large-scale organization emerges from interacting components remains a central challenge across physics, biology, and artificial intelligence. In this work, we establish a spectral framework that both identifies and reproduces emergent structure using random matrix theory (RMT) as a quantitative baseline. We employ the ratio of level spacings ( $r$ -statistic) and its variance as unfolding-free observables to characterize spectral organization across scales. Numerical experiments on classical random matrix ensembles (GOE, GUE, and Poisson) demonstrate immediate universality: spectral statistics converge at small system sizes ( $N \ge 100$ ), remain stable across scales, and exhibit no intermediate mesoscopic regime. This confirms that RMT provides a null model devoid of emergent behavior. We then introduce a minimal correlated random matrix model with three competing components: noise, structured correlations, and a low-rank signal tuned to the Baik–Ben Arous–Péché (BBP) transition. This controlled setting reveals a genuine mesoscopic regime characterized by a finite-size peak in the variance of the edge statistic $\mathrm {Var}(r)$ , accompanied by persistent restructuring of the bulk spectrum. The effect is maximized at criticality and disappears asymptotically, establishing emergence as a finite-size phenomenon driven by competing scales. These results provide a concrete and operational definition of emergence as a deviation from spectral universality. The proposed framework transforms spectral analysis into a diagnostic tool: stable RMT-like behavior indicates absence of structure, while mesoscopic variance peaks signal partial organization, criticality, and ongoing reorganization. This approach is general and directly applicable to complex systems, including neural networks, biological data, and correlated physical systems, where it offers a new lens for detecting structure formation and dynamical transitions.
Förderung
- Higher Colleges of Technology, United Arab Emirates
Zitieren
APA
Siwe, A. T. (2026). Spectral Diagnostics of Emergence: A Random Matrix Baseline and a Mesoscopic Regime From Competing Scales. IEEE Access, 14, 83457-83470. https://doi.org/10.1109/access.2026.3696391
Vancouver
Siwe AT. Spectral Diagnostics of Emergence: A Random Matrix Baseline and a Mesoscopic Regime From Competing Scales. IEEE Access. 2026;14:83457-83470. doi:10.1109/access.2026.3696391
BibTeX
@article{Siwe2026,
author = {Siwe, Alain Tcheukam},
title = {Spectral Diagnostics of Emergence: A Random Matrix Baseline and a Mesoscopic Regime From Competing Scales},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {83457--83470},
doi = {10.1109/access.2026.3696391}
}
EUVIMED